Plan for Analysis

Plot F0 and 95% confidence intervals on aggregate.
Plot F0 and 95% confidence intervals by gender, by noise levels, by vowels, by vowel and noise levels, by environment (isolation or in a carrier sentence) and also create a regression line for each categorization.
Summarize findings & email Yan Other notes Also leave one speaker out and try classification. If the reslt is good we can use it to predict tone produced on what kind of level (leave the noise level aside). The noise shouldn’t matter much at this stage. (18 levels?) look at similarity of tones (using clustering).

# Read in all the F0 contour info. All of these f0.mat files have been downsampled
# in MatLab during linear interpolation (see getf0.mat in the matlab codes)

# matlab code:
# myFolder = '\\client\h$\Desktop\ProsodyPro\m-3-78\channel1';
# myFiles = dir(fullfile(myFolder,'*.wav')); %gets all wav files
# 
# for k = 1:length(myFiles)
#   baseFileName = myFiles(k).name;
#   fullFileName = fullfile(myFolder, baseFileName);
#   fprintf(1, 'Now reading %s\n', fullFileName);
#   [y, Fs] = audioread(fullFileName);
#   [f0, ~] = pitchRocco(y, Fs);
#   i = 1:length(f0);
#   i_new =  linspace(min(i), max(i), 20);
#   f0_downsampled = interp1(i, f0, i_new, 'linear');  
#   fid= fopen(fullfile(myFolder, 'F0.mat'),'a');
#   fprintf(fid, '%s ', baseFileName);
#   fprintf(fid, '%f ', f0_downsampled);
#   fprintf(fid,'\n');
#   fclose(fid);
# end

# Step 1: Read in all F0.mat files into f0Files and assign a name for each

f0Files <- lapply(Sys.glob("*/channel1/F0.mat"), read.table)
length(f0Files)

## [1] 12

f0_f_1_78 <- as.data.frame(f0Files[1])
f0_f_1_90 <- as.data.frame(f0Files[2])
f0_f_1_q <- as.data.frame(f0Files[3])
f0_m_1_78 <- as.data.frame(f0Files[4])
f0_m_1_90 <- as.data.frame(f0Files[5])
f0_m_1_q <- as.data.frame(f0Files[6])
f0_m_2_78 <- as.data.frame(f0Files[7])
f0_m_2_90 <- as.data.frame(f0Files[8])
f0_m_2_q <- as.data.frame(f0Files[9])
f0_m_3_78 <- as.data.frame(f0Files[10])
f0_m_3_90 <- as.data.frame(f0Files[11])  
f0_m_3_q <- as.data.frame(f0Files[12])

# Step 2: Add column names for all dataframes.

numbers <- 1:20
cols <- c("sound.name",numbers)
colnames(f0_f_1_78) <- cols
colnames(f0_f_1_90) <- cols
colnames(f0_f_1_q) <- cols
colnames(f0_m_1_78) <- cols
colnames(f0_m_1_90) <- cols
colnames(f0_m_1_q) <- cols
colnames(f0_m_2_78) <- cols
colnames(f0_m_2_90) <- cols
colnames(f0_m_2_q) <- cols
colnames(f0_m_3_78) <- cols
colnames(f0_m_3_90) <- cols
colnames(f0_m_3_q) <- cols


# Assigning noise level
f0_f_1_78$noise = 78
f0_f_1_90$noise = 90
f0_f_1_q$noise = 0

f0_m_1_78$noise = 78
f0_m_1_90$noise = 90
f0_m_1_q$noise = 0

f0_m_2_78$noise = 78
f0_m_2_90$noise = 90
f0_m_2_q$noise = 0

f0_m_3_78$noise = 78
f0_m_3_90$noise = 90
f0_m_3_q$noise = 0


# Assigning gender variable (0 for female and 1 for male)
f0_f_1_78$gender = 0
f0_f_1_90$gender = 0
f0_f_1_q$gender = 0

f0_m_1_78$gender = 1
f0_m_1_90$gender = 1
f0_m_1_q$gender = 1

f0_m_2_78$gender = 1
f0_m_2_90$gender = 1
f0_m_2_q$gender = 1

f0_m_3_78$gender = 1
f0_m_3_90$gender = 1
f0_m_3_q$gender = 1


### Concatenate all dataframes
f0_reports <- rbind(f0_f_1_78, f0_f_1_90, f0_f_1_q,
f0_m_1_78, f0_m_1_90, f0_m_1_q,
f0_m_2_78, f0_m_2_90, f0_m_2_q,
f0_m_3_78, f0_m_3_90, f0_m_3_q)

According to the report, this dataframe has 3786 records. In total, from 4 speakers, we have 3786 sound segments to analyze.

# Assign tone values.

# Extracting a substring that contains only the syllable names.
nameswithoutwav <- sapply(strsplit(f0_reports[,1], split=".", fixed=TRUE), "[", 1)
f0_reports$syllable.names <- sapply(strsplit(nameswithoutwav, split="_", fixed=TRUE), "[", 1)

f0_reports$tone <- ifelse(grepl("a", f0_reports$syllable.name, ignore.case=T), "A1",
ifelse(grepl("^tát", f0_reports$syllable.name, ignore.case=T), "D1",
ifelse(grepl("^tạt", f0_reports$syllable.name, ignore.case=T), "D2",
ifelse(grepl("^tết", f0_reports$syllable.name, ignore.case=T), "D1",
ifelse(grepl("^tệt", f0_reports$syllable.name, ignore.case=T), "D2",
ifelse(grepl("^tút", f0_reports$syllable.name, ignore.case=T), "D1",
ifelse(grepl("^tụt", f0_reports$syllable.name, ignore.case=T), "D2",
ifelse(grepl("à", f0_reports$syllable.name, ignore.case=T), "A2",
ifelse(grepl("á", f0_reports$syllable.name, ignore.case=T), "B1",
ifelse(grepl("ả", f0_reports$syllable.name, ignore.case=T), "C1",
ifelse(grepl("ã", f0_reports$syllable.name, ignore.case=T), "C2",
ifelse(grepl("ạ", f0_reports$syllable.name, ignore.case=T), "B2",
ifelse(grepl("ê", f0_reports$syllable.name, ignore.case=T), "A1",
ifelse(grepl("ề", f0_reports$syllable.name, ignore.case=T), "A2",
ifelse(grepl("ế", f0_reports$syllable.name, ignore.case=T), "B1",
ifelse(grepl("ể", f0_reports$syllable.name, ignore.case=T), "C1",
ifelse(grepl("ễ", f0_reports$syllable.name, ignore.case=T), "C2",
ifelse(grepl("ệ", f0_reports$syllable.name, ignore.case=T), "B2",
ifelse(grepl("u", f0_reports$syllable.name, ignore.case=T), "A1",
ifelse(grepl("ù", f0_reports$syllable.name, ignore.case=T), "A2",
ifelse(grepl("ú", f0_reports$syllable.name, ignore.case=T), "B1",
ifelse(grepl("ủ", f0_reports$syllable.name, ignore.case=T), "C1",
ifelse(grepl("ũ", f0_reports$syllable.name, ignore.case=T), "C2",
ifelse(grepl("ụ", f0_reports$syllable.name, ignore.case=T), "B2",
ifelse(grepl("ộ", f0_reports$syllable.name, ignore.case=T), "B2","NA")))))))))))))))))))))))))

# Assigning if the token is single (1) or not (0). Single tokens were produced in isolation.
# Otherwise they were produced in carrier sentences.
f0_reports$single <- ifelse(grepl("single", f0_reports$sound.name), 1, 0)

# Convert categorical variables to factor levels.
f0_reports$gender <- as.factor(f0_reports$gender)
f0_reports$single <- as.factor(f0_reports$single)
f0_reports$tone <- as.factor(f0_reports$tone)
f0_reports$noise <- as.factor(f0_reports$noise)

summary(f0_reports)

##   sound.name              1                2               3        
##  Length:3786        Min.   :  0.00   Min.   :  0.0   Min.   :  0.0  
##  Class :character   1st Qu.:  0.00   1st Qu.:117.4   1st Qu.:125.1  
##  Mode  :character   Median :115.91   Median :142.4   Median :147.2  
##                     Mean   : 95.73   Mean   :140.6   Mean   :152.6  
##                     3rd Qu.:148.73   3rd Qu.:169.8   3rd Qu.:175.9  
##                     Max.   :376.52   Max.   :353.6   Max.   :369.0  
##                                                                     
##        4               5               6               7        
##  Min.   :  0.0   Min.   :  0.0   Min.   :  0.0   Min.   :  0.0  
##  1st Qu.:125.9   1st Qu.:124.7   1st Qu.:122.4   1st Qu.:119.8  
##  Median :147.1   Median :146.9   Median :145.0   Median :143.4  
##  Mean   :154.6   Mean   :153.9   Mean   :151.7   Mean   :149.1  
##  3rd Qu.:176.9   3rd Qu.:177.5   3rd Qu.:177.0   3rd Qu.:174.8  
##  Max.   :322.7   Max.   :308.9   Max.   :304.4   Max.   :306.9  
##                                                                 
##        8               9               10              11       
##  Min.   :  0.0   Min.   :  0.0   Min.   :  0.0   Min.   :  0.0  
##  1st Qu.:117.5   1st Qu.:115.8   1st Qu.:114.2   1st Qu.:113.0  
##  Median :142.3   Median :141.7   Median :140.9   Median :139.3  
##  Mean   :147.5   Mean   :147.0   Mean   :146.3   Mean   :145.7  
##  3rd Qu.:173.8   3rd Qu.:174.0   3rd Qu.:174.7   3rd Qu.:176.3  
##  Max.   :307.6   Max.   :327.2   Max.   :308.7   Max.   :312.0  
##                                                                 
##        12              13              14              15       
##  Min.   :  0.0   Min.   :  0.0   Min.   :  0.0   Min.   :  0.0  
##  1st Qu.:111.0   1st Qu.:109.6   1st Qu.:108.5   1st Qu.:107.5  
##  Median :139.1   Median :139.8   Median :139.9   Median :140.6  
##  Mean   :144.3   Mean   :144.6   Mean   :144.7   Mean   :144.9  
##  3rd Qu.:177.7   3rd Qu.:180.2   3rd Qu.:183.3   3rd Qu.:185.6  
##  Max.   :319.8   Max.   :327.8   Max.   :336.7   Max.   :356.8  
##                                                                 
##        16              17              18              19       
##  Min.   :  0.0   Min.   :  0.0   Min.   :  0.0   Min.   :  0.0  
##  1st Qu.:107.3   1st Qu.:105.6   1st Qu.:102.3   1st Qu.: 96.9  
##  Median :141.2   Median :138.4   Median :134.2   Median :127.6  
##  Mean   :145.2   Mean   :142.8   Mean   :137.2   Mean   :129.0  
##  3rd Qu.:186.7   3rd Qu.:184.8   3rd Qu.:180.4   3rd Qu.:170.6  
##  Max.   :398.6   Max.   :406.8   Max.   :413.7   Max.   :377.8  
##                                                                 
##        20         noise     gender   syllable.names          tone     single  
##  Min.   :  0.00   0 :1260   0: 947   Length:3786        B2     :624   0:1893  
##  1st Qu.:  0.00   78:1263   1:2839   Class :character   A2     :575   1:1893  
##  Median : 97.47   90:1263            Mode  :character   B1     :575           
##  Mean   : 83.49                                         C1     :575           
##  3rd Qu.:135.47                                         C2     :575           
##  Max.   :343.06                                         A1     :574           
##                                                         (Other):288

Aggregate Tone Analysis

F0 Plots

# Note that I only plot the 2nd to 19th sampled points.
# Extract F0 of all tones A1 (574 instances)
par(mfrow=c(2,4))
A1 <- data.matrix(f0_reports[f0_reports$tone=="A1",-c(1,c(22:26))])
plot(colMeans(A1)[2:19], type="l", xlim=c(1, 18), ylim=c(60, 200))

# # Do the same for the other tones
A2 <- data.matrix(f0_reports[f0_reports$tone=="A2",-c(1,c(22:26))])
plot(colMeans(A2)[2:19], type="l", xlim=c(1, 18), ylim=c(60, 200))

B1 <- data.matrix(f0_reports[f0_reports$tone=="B1",-c(1,c(22:26))])
plot(colMeans(B1)[2:19], type="l", xlim=c(1, 18), ylim=c(60, 200))

B2 <- data.matrix(f0_reports[f0_reports$tone=="B2",-c(1,c(22:26))])
plot(colMeans(B2)[2:19], type="l", xlim=c(1, 18), ylim=c(60, 250))

C1 <- data.matrix(f0_reports[f0_reports$tone=="C1",-c(1,c(22:26))])
plot(colMeans(C1)[2:19], type="l", xlim=c(1, 18), ylim=c(60, 250))

C2 <- data.matrix(f0_reports[f0_reports$tone=="C2",-c(1,c(22:26))])
plot(colMeans(C2)[2:19], type="l", xlim=c(1, 18), ylim=c(60, 250))

D1 <- data.matrix(f0_reports[f0_reports$tone=="D1",-c(1,c(22:26))])
plot(colMeans(D1)[2:19], type="l", xlim=c(1, 18), ylim=c(60, 250))

D2 <- data.matrix(f0_reports[f0_reports$tone=="D2",-c(1,c(22:26))])
plot(colMeans(D2)[2:19], type="l", xlim=c(1, 18), ylim=c(60, 250))

F0 contours on the same plot.

plot(colMeans(A1)[2:19], type="b", xlim=c(1, 18), ylim=c(60, 200))
lines(colMeans(A2)[2:19],col="green", type="b", pch=19)
lines(colMeans(B1)[2:19],col="red",  type="b", pch=19)
lines(colMeans(B2)[2:19],col="purple", type="b",  pch=19)
lines(colMeans(C1)[2:19],col="blue",  type="b", pch=19)
lines(colMeans(C2)[2:19],col="orange",  type="b", pch=19)

# Add a legend
legend(1, 100, legend=c("A1", "A2", "B1", "B2", "C1", "C2"),
       col=c("black", "green", "red", "purple", "blue", "orange"), lty=1:2, cex=0.7)

Plot B1, B2, D1, D2 together.

plot(colMeans(B1)[2:19], type="b", col="blue", xlim=c(1, 18), ylim=c(60, 200))
lines(colMeans(B2)[2:19],col="purple", type="b",  pch=19)
lines(colMeans(D1)[2:19],col="pink",  type="b", pch=19)  ## D1 has a very strange contour.
lines(colMeans(D2)[2:19],col="black",  type="b", pch=19)

# Add a legend
legend(1, 100, legend=c("B1", "B2", "D1", "D2"),
       col=c("blue", "orange", "pink", "black"), lty=1:2, cex=0.7)

Estimate Regression Lines for Each Tone on Aggregate.

regression_report <- function(tone) {
  lm_tone <- lm(colMeans(tone)[2:19]~ c(1:18))
  plot(colMeans(tone)[2:19], pch = 16, cex = 1.3, xlim=c(1, 18), ylim=c(60, 250))
  abline(lm(colMeans(tone)[2:19] ~ c(1:18)))
  return(lm_tone)
}

## Call regression_report on all tones.
regression_report(A1)

## 
## Call:
## lm(formula = colMeans(tone)[2:19] ~ c(1:18))
## 
## Coefficients:
## (Intercept)      c(1:18)  
##    167.5833      -0.6218

regression_report(A2)

## 
## Call:
## lm(formula = colMeans(tone)[2:19] ~ c(1:18))
## 
## Coefficients:
## (Intercept)      c(1:18)  
##     144.641       -1.454

regression_report(B1)

## 
## Call:
## lm(formula = colMeans(tone)[2:19] ~ c(1:18))
## 
## Coefficients:
## (Intercept)      c(1:18)  
##     122.987        3.414

regression_report(B2)

## 
## Call:
## lm(formula = colMeans(tone)[2:19] ~ c(1:18))
## 
## Coefficients:
## (Intercept)      c(1:18)  
##      176.53        -5.22

regression_report(C1)

## 
## Call:
## lm(formula = colMeans(tone)[2:19] ~ c(1:18))
## 
## Coefficients:
## (Intercept)      c(1:18)  
##     151.462       -2.647

regression_report(C2)

## 
## Call:
## lm(formula = colMeans(tone)[2:19] ~ c(1:18))
## 
## Coefficients:
## (Intercept)      c(1:18)  
##     153.883        1.916

regression_report(D1)

## 
## Call:
## lm(formula = colMeans(tone)[2:19] ~ c(1:18))
## 
## Coefficients:
## (Intercept)      c(1:18)  
##    164.3336       0.8907

regression_report(D2)

## 
## Call:
## lm(formula = colMeans(tone)[2:19] ~ c(1:18))
## 
## Coefficients:
## (Intercept)      c(1:18)  
##     148.707       -2.237

Tone Contour Analysis on Different Noise Levels (Separately by Single vs. Tokens in Carriers)

# Plot F0 contours according to different levels

filter_f0 <- function(tone, noise, s) {
  if (s==TRUE) {
    tone_matrix <- data.matrix(f0_reports[f0_reports$tone==tone & f0_reports$noise==noise & f0_reports$single=="1",-c(1,c(22:26))])
    return(tone_matrix)
  } else {
    tone_matrix <- data.matrix(f0_reports[f0_reports$tone==tone & f0_reports$noise==noise & f0_reports$single=="0",-c(1,c(22:26))])
    return(tone_matrix)
  }
}

plot_f0_by_noise <- function(tone, s) {
  if (s==TRUE) {
    tone_quiet <- data.matrix(f0_reports[f0_reports$tone==tone & f0_reports$noise=="0" & f0_reports$single=="1",-c(1,c(22:26))])
  tone_78 <- data.matrix(f0_reports[f0_reports$tone==tone & f0_reports$noise=="78" & f0_reports$single=="1",-c(1,c(22:26))])
  tone_90 <- data.matrix(f0_reports[f0_reports$tone==tone & f0_reports$noise=="90" & f0_reports$single=="1",-c(1,c(22:26))])
  plot(colMeans(tone_quiet)[2:19], type="l", xlim=c(1, 18), ylim=c(70, 250))
  lines(colMeans(tone_78)[2:19],col="green")
  lines(colMeans(tone_90)[2:19],col="red")
  # Find regression coefficients and return them.
  c(summary(regression_report(tone_quiet))$coefficients[2,1], summary(regression_report(tone_78))$coefficients[2,1], summary(regression_report(tone_90))$coefficients[2,1])
  } else {
    tone_quiet <- data.matrix(f0_reports[f0_reports$tone==tone & f0_reports$noise=="0" & f0_reports$single=="0",-c(1,c(22:26))])
  tone_78 <- data.matrix(f0_reports[f0_reports$tone==tone & f0_reports$noise=="78" & f0_reports$single=="0",-c(1,c(22:26))])
  tone_90 <- data.matrix(f0_reports[f0_reports$tone==tone & f0_reports$noise=="90" & f0_reports$single=="0",-c(1,c(22:26))])
  plot(colMeans(tone_quiet)[2:19], type="l", xlim=c(1, 18), ylim=c(70, 250))
  lines(colMeans(tone_78)[2:19],col="green")
  lines(colMeans(tone_90)[2:19],col="red")
  c(summary(regression_report(tone_quiet))$coefficients[2,1], summary(regression_report(tone_78))$coefficients[2,1], summary(regression_report(tone_90))$coefficients[2,1])
  }
}

coefficient_reports <- function(tone_matrix) {
  lm_model <- lm(colMeans(tone_matrix)[2:19] ~ c(1:18))
  summary(lm_model)$coefficients[2,1]
}

F0 and regression line for tone A1, single tokens.

plot_f0_by_noise("A1", TRUE)

## [1] -0.5672417 -0.8400165 -0.5233609

F0 and regression line for tone A2, single tokens.

plot_f0_by_noise("A2",TRUE)

## [1] -1.378635 -1.759775 -1.540135

F0 and regression line for tone B1, single tokens.

plot_f0_by_noise("B1", TRUE)

## [1] 2.585417 4.653510 5.201106

F0 and regression line for tone B2, single tokens.

plot_f0_by_noise("B2",TRUE)

## [1] -3.416345 -6.636740 -6.309881

F0 and regression line for tone C1, single tokens.

plot_f0_by_noise("C1",TRUE)

## [1] -1.254551 -1.406069 -1.978960

###F0 and regression line for tone C2, single tokens.

plot_f0_by_noise("C2", TRUE)

## [1] 2.5676060 1.8890026 0.8420454

F0 and regression line for tone D1, single tokens.

plot_f0_by_noise("D1",TRUE)

## [1] -0.3939798  3.5934800  1.1236231

F0 and regression line for tone D2, single tokens.

plot_f0_by_noise("D2", TRUE)

## [1] -1.457687 -3.538611 -2.323780

F0 and regression line for tone A1, tokens in carriers.

plot_f0_by_noise("A1", FALSE)

## [1] -0.6754738 -0.5120815 -0.6169642

F0 and regression line for tone A2, tokens in carriers.

plot_f0_by_noise("A2",FALSE)

## [1] -1.133775 -1.282178 -1.621849

F0 and regression line for tone B1, tokens in carriers.

plot_f0_by_noise("B1", FALSE)

## [1] 2.212407 2.938129 2.887313

F0 and regression line for tone B2, tokens in carriers.

plot_f0_by_noise("B2",FALSE)

## [1] -4.002273 -4.932588 -6.022167

F0 and regression line for tone C1, tokens in carriers.

plot_f0_by_noise("C1",FALSE)

## [1] -2.854671 -3.511772 -4.884935

F0 and regression line for tone C2, tokens in carriers.

plot_f0_by_noise("C2", FALSE)

## [1] 2.584509 2.157254 1.460962

F0 and regression line for tone D1, tokens in carriers.

plot_f0_by_noise("D1",FALSE)

## [1]  0.3983035 -0.0583307  0.6809495

F0 and regression line for tone D2, tokens in carriers.

plot_f0_by_noise("D2", FALSE)

## [1] -2.531688 -2.139725 -1.431408

All coefficients together in a table.

# Table of tone * noise (20x3)
tab <- matrix(,nrow=16, ncol=3, byrow=TRUE)
colnames(tab) <- c('quiet','noise 78','noise 90')
rownames(tab) <- c('A1 single','A2 single','B1 single', 'B2 single',
                   'C1 single','C2 single','D1 single', 'D2 single',
                   'A1 carrier','A2 carrier','B1 carrier', 'B2 carrier',
                   'C1 carrier','C2 carrier','D1 carrier', 'D2 carrier')
tab <- as.table(tab)
tab[1,1] = coefficient_reports(filter_f0("A1", "0", TRUE))
tab[1,2] = coefficient_reports(filter_f0("A1", "78", TRUE))
tab[1,3] = coefficient_reports(filter_f0("A1", "90", TRUE))

tab[2,1] = coefficient_reports(filter_f0("A2", "0", TRUE))
tab[2,2] = coefficient_reports(filter_f0("A2", "78", TRUE))
tab[2,3] = coefficient_reports(filter_f0("A2", "90", TRUE))

tab[3,1] = coefficient_reports(filter_f0("B1", "0", TRUE))
tab[3,2] = coefficient_reports(filter_f0("B1", "78", TRUE))
tab[3,3] = coefficient_reports(filter_f0("B1", "90", TRUE))

tab[4,1] = coefficient_reports(filter_f0("B2", "0", TRUE))
tab[4,2] = coefficient_reports(filter_f0("B2", "78", TRUE))
tab[4,3] = coefficient_reports(filter_f0("B2", "90", TRUE))

tab[5,1] = coefficient_reports(filter_f0("C1", "0", TRUE))
tab[5,2] = coefficient_reports(filter_f0("C1", "78", TRUE))
tab[5,3] = coefficient_reports(filter_f0("C1", "90", TRUE))

tab[6,1] = coefficient_reports(filter_f0("C2", "0", TRUE))
tab[6,2] = coefficient_reports(filter_f0("C2", "78", TRUE))
tab[6,3] = coefficient_reports(filter_f0("C2", "90", TRUE))

tab[7,1] = coefficient_reports(filter_f0("D1", "0", TRUE))
tab[7,2] = coefficient_reports(filter_f0("D1", "78", TRUE))
tab[7,3] = coefficient_reports(filter_f0("D1", "90", TRUE))

tab[8,1] = coefficient_reports(filter_f0("D2", "0", TRUE))
tab[8,2] = coefficient_reports(filter_f0("D2", "78", TRUE))
tab[8,3] = coefficient_reports(filter_f0("D2", "90", TRUE))

# carrier
tab[9,1] = coefficient_reports(filter_f0("A1", "0", FALSE))
tab[9,2] = coefficient_reports(filter_f0("A1", "78", FALSE))
tab[9,3] = coefficient_reports(filter_f0("A1", "90", FALSE))

tab[10,1] = coefficient_reports(filter_f0("A2", "0", FALSE))
tab[10,2] = coefficient_reports(filter_f0("A2", "78", FALSE))
tab[10,3] = coefficient_reports(filter_f0("A2", "90", FALSE))

tab[11,1] = coefficient_reports(filter_f0("B1", "0", FALSE))
tab[11,2] = coefficient_reports(filter_f0("B1", "78", FALSE))
tab[11,3] = coefficient_reports(filter_f0("B1", "90", FALSE))

tab[12,1] = coefficient_reports(filter_f0("B2", "0", FALSE))
tab[12,2] = coefficient_reports(filter_f0("B2", "78", FALSE))
tab[12,3] = coefficient_reports(filter_f0("B2", "90", FALSE))

tab[13,1] = coefficient_reports(filter_f0("C1", "0", FALSE))
tab[13,2] = coefficient_reports(filter_f0("C1", "78", FALSE))
tab[13,3] = coefficient_reports(filter_f0("C1", "90", FALSE))

tab[14,1] = coefficient_reports(filter_f0("C2", "0", FALSE))
tab[14,2] = coefficient_reports(filter_f0("C2", "78", FALSE))
tab[14,3] = coefficient_reports(filter_f0("C2", "90", FALSE))

tab[15,1] = coefficient_reports(filter_f0("D1", "0", FALSE))
tab[15,2] = coefficient_reports(filter_f0("D1", "78", FALSE))
tab[15,3] = coefficient_reports(filter_f0("D1", "90", FALSE))

tab[16,1] = coefficient_reports(filter_f0("D2", "0", FALSE))
tab[16,2] = coefficient_reports(filter_f0("D2", "78", FALSE))
tab[16,3] = coefficient_reports(filter_f0("D2", "90", FALSE))

tab

##                 quiet   noise 78   noise 90
## A1 single  -0.5672417 -0.8400165 -0.5233609
## A2 single  -1.3786352 -1.7597751 -1.5401346
## B1 single   2.5854175  4.6535103  5.2011062
## B2 single  -3.4163446 -6.6367399 -6.3098807
## C1 single  -1.2545513 -1.4060692 -1.9789597
## C2 single   2.5676060  1.8890026  0.8420454
## D1 single  -0.3939798  3.5934800  1.1236231
## D2 single  -1.4576871 -3.5386107 -2.3237797
## A1 carrier -0.6754738 -0.5120815 -0.6169642
## A2 carrier -1.1337752 -1.2821783 -1.6218491
## B1 carrier  2.2124070  2.9381288  2.8873133
## B2 carrier -4.0022734 -4.9325882 -6.0221675
## C1 carrier -2.8546710 -3.5117715 -4.8849348
## C2 carrier  2.5845091  2.1572536  1.4609616
## D1 carrier  0.3983035 -0.0583307  0.6809495
## D2 carrier -2.5316879 -2.1397249 -1.4314081

Tone Contour Analysis on Different Noise Levels on Vowel A.

# Add vowel annotation.

f0_reports$vowel <- ifelse(grepl("a", f0_reports$syllable.name, ignore.case=T), "A",
ifelse(grepl("à", f0_reports$syllable.name, ignore.case=T), "A",
ifelse(grepl("á", f0_reports$syllable.name, ignore.case=T), "A",
ifelse(grepl("ả", f0_reports$syllable.name, ignore.case=T), "A",
ifelse(grepl("ã", f0_reports$syllable.name, ignore.case=T), "A",
ifelse(grepl("ạ", f0_reports$syllable.name, ignore.case=T), "A",
ifelse(grepl("ê", f0_reports$syllable.name, ignore.case=T), "E",
ifelse(grepl("ề", f0_reports$syllable.name, ignore.case=T), "E",
ifelse(grepl("ế", f0_reports$syllable.name, ignore.case=T), "E",
ifelse(grepl("ể", f0_reports$syllable.name, ignore.case=T), "E",
ifelse(grepl("ễ", f0_reports$syllable.name, ignore.case=T), "E",
ifelse(grepl("ệ", f0_reports$syllable.name, ignore.case=T), "E",
ifelse(grepl("u", f0_reports$syllable.name, ignore.case=T), "U",
ifelse(grepl("ù", f0_reports$syllable.name, ignore.case=T), "U",
ifelse(grepl("ú", f0_reports$syllable.name, ignore.case=T), "U",
ifelse(grepl("ủ", f0_reports$syllable.name, ignore.case=T), "U",
ifelse(grepl("ũ", f0_reports$syllable.name, ignore.case=T), "U",
ifelse(grepl("ụ", f0_reports$syllable.name, ignore.case=T), "U",
ifelse(grepl("ộ", f0_reports$syllable.name, ignore.case=T), "O","NA")))))))))))))))))))

# Convert vowel types to a factor variable
f0_reports$vowel <- as.factor(f0_reports$vowel)
head(f0_reports, 30)

# Plot F0 contour by vowels (not distinguishing single tokens or tokens in carriers.)

plot_f0_by_vowels <- function(tone, vowel) {
  if (vowel=="A") {
    tone_quiet <- data.matrix(f0_reports[f0_reports$tone==tone & f0_reports$noise=="0" & f0_reports$vowel=="A",-c(1,c(22:26))])
  tone_78 <- data.matrix(f0_reports[f0_reports$tone==tone & f0_reports$noise=="78" & f0_reports$vowel=="A",-c(1,c(22:26))])
  tone_90 <- data.matrix(f0_reports[f0_reports$tone==tone & f0_reports$noise=="90" & f0_reports$vowel=="A",-c(1,c(22:26))])
  plot(colMeans(tone_quiet)[2:19], type="l", xlim=c(1, 18), ylim=c(70, 250))
  lines(colMeans(tone_78)[2:19],col="green")
  lines(colMeans(tone_90)[2:19],col="red")
  # Find regression coefficients and return them.
  c(summary(regression_report(tone_quiet))$coefficients[2,1], summary(regression_report(tone_78))$coefficients[2,1], summary(regression_report(tone_90))$coefficients[2,1])
  } else if (vowel=="E") {
    tone_quiet <- data.matrix(f0_reports[f0_reports$tone==tone & f0_reports$noise=="0" & f0_reports$vowel=="E",-c(1,c(22:26))])
  tone_78 <- data.matrix(f0_reports[f0_reports$tone==tone & f0_reports$noise=="78" & f0_reports$vowel=="E",-c(1,c(22:26))])
  tone_90 <- data.matrix(f0_reports[f0_reports$tone==tone & f0_reports$noise=="90" & f0_reports$vowel=="E",-c(1,c(22:26))])
  plot(colMeans(tone_quiet)[2:19], type="l", xlim=c(1, 18), ylim=c(70, 250))
  lines(colMeans(tone_78)[2:19],col="green")
  lines(colMeans(tone_90)[2:19],col="red")
  # Find regression coefficients and return them.
  c(summary(regression_report(tone_quiet))$coefficients[2,1], summary(regression_report(tone_78))$coefficients[2,1], summary(regression_report(tone_90))$coefficients[2,1])
  } else {
  tone_quiet <- data.matrix(f0_reports[f0_reports$tone==tone & f0_reports$noise=="0" & f0_reports$vowel=="U",-c(1,c(22:26))])
  tone_78 <- data.matrix(f0_reports[f0_reports$tone==tone & f0_reports$noise=="78" & f0_reports$vowel=="U",-c(1,c(22:26))])
  tone_90 <- data.matrix(f0_reports[f0_reports$tone==tone & f0_reports$noise=="90" & f0_reports$vowel=="U",-c(1,c(22:26))])
  plot(colMeans(tone_quiet)[2:19], type="l", xlim=c(1, 18), ylim=c(70, 250))
  lines(colMeans(tone_78)[2:19],col="green")
  lines(colMeans(tone_90)[2:19],col="red")
  # Find regression coefficients and return them.
  c(summary(regression_report(tone_quiet))$coefficients[2,1], summary(regression_report(tone_78))$coefficients[2,1], summary(regression_report(tone_90))$coefficients[2,1])
  }
}

# Call functions

Tone A1 on A

plot_f0_by_vowels("A1", "A")

## [1] -0.2878660 -0.2424775 -0.3018372

Tone A1 on E

plot_f0_by_vowels("A1", "E")

## [1] -0.2925254 -0.8486117 -0.6995807

Tone A1 on U

plot_f0_by_vowels("A1", "U")

## [1] -1.2836820 -0.9401882 -0.7090698

Tone A2 on A

plot_f0_by_vowels("A2", "A")

## [1] -0.9418938 -1.2171543 -1.0407443

Tone A2 on E

plot_f0_by_vowels("A2", "E")

## [1] -1.191150 -1.487463 -1.654423

Tone A2 on U

plot_f0_by_vowels("A2", "U")

## [1] -1.629644 -1.855327 -2.038729

Tone B1 on A

plot_f0_by_vowels("B1", "A")

## [1] 3.299303 4.444733 4.265991

Tone B1 on E

plot_f0_by_vowels("B1", "E")

## [1] 3.593280 4.729655 4.469133

Tone B1 on U

plot_f0_by_vowels("B1", "U")

## [1] 0.2679428 2.2130707 3.3975057

Tone B2 on A

plot_f0_by_vowels("B2", "A")

## [1] -3.683201 -5.224309 -6.272141

Tone B2 on E

plot_f0_by_vowels("B2", "E")

## [1] -3.547011 -5.916156 -5.927092

Tone B2 on U

plot_f0_by_vowels("B2", "U")

## [1] -4.080879 -6.072111 -6.298078

Tone C1 on A

plot_f0_by_vowels("C1", "A")

## [1] -2.136312 -2.063681 -2.753528

Tone C1 on E

plot_f0_by_vowels("C1", "E")

## [1] -2.071855 -2.160343 -3.665938

Tone C1 on U

plot_f0_by_vowels("C1", "U")

## [1] -1.955666 -3.130111 -3.876376

Tone C2 on A

plot_f0_by_vowels("C2", "A")

## [1] 3.010333 3.027028 2.394154

Tone C2 on E

plot_f0_by_vowels("C2", "E")

## [1] 3.102830 1.836030 1.096396

Tone C2 on U

plot_f0_by_vowels("C2", "U")

## [1]  1.62192705  1.20632671 -0.03604034

Tone D1 on A

plot_f0_by_vowels("D1", "A")

## [1] 1.946698 1.572546 2.110463

Tone D1 on E

plot_f0_by_vowels("D1", "E")

## [1] 0.6420803 4.6188919 2.2607038

Tone D1 on U

plot_f0_by_vowels("D1", "U")

## [1] -2.5822926 -0.8887144 -1.6643078

Tone D2 on A

plot_f0_by_vowels("D2", "A")

## [1] -1.411010 -3.515305 -2.333420

Tone D2 on E

plot_f0_by_vowels("D2", "E")

## [1] -0.5778161 -2.1378543 -1.1907024

Tone D2 on U

plot_f0_by_vowels("D2", "U")

## [1] -3.995236 -2.864344 -2.108659

Slope coefficients in a table

Plot F0 contours according to different levels

filter_f0 <- function(tone, vowel, noise) {
    tone_matrix <- data.matrix(f0_reports[f0_reports$tone==tone & f0_reports$noise==noise & f0_reports$vowel==vowel,-c(1,c(22:26))])
    return(tone_matrix)
}

# Table of tone * noise (24x3)
tab_vowel <- matrix(,nrow=24, ncol=3, byrow=TRUE)
colnames(tab_vowel) <- c('quiet','noise 78','noise 90')
rownames(tab_vowel) <- c('A1 A','A2 A','B1 A', 'B2 A',
                   'C1 A','C2 A','D1 A', 'D2 A',
                   'A1 E','A2 E','B1 E', 'B2 E',
                   'C1 E','C2 E','D1 E', 'D2 E',
                   'A1 U','A2 U','B1 U', 'B2 U',
                   'C1 U','C2 U','D1 U', 'D2 U')
tab_vowel <- as.table(tab_vowel)
tab_vowel[1,1] = coefficient_reports(filter_f0("A1", "A", "0"))
tab_vowel[1,2] = coefficient_reports(filter_f0("A1", "A", "78"))
tab_vowel[1,3] = coefficient_reports(filter_f0("A1", "A", "90"))

tab_vowel[2,1] = coefficient_reports(filter_f0("A2", "A", "0"))
tab_vowel[2,2] = coefficient_reports(filter_f0("A2", "A", "78"))
tab_vowel[2,3] = coefficient_reports(filter_f0("A2", "A", "90"))

tab_vowel[3,1] = coefficient_reports(filter_f0("B1", "A", "0"))
tab_vowel[3,2] = coefficient_reports(filter_f0("B1", "A", "78"))
tab_vowel[3,3] = coefficient_reports(filter_f0("B1", "A", "90"))

tab_vowel[4,1] = coefficient_reports(filter_f0("B2", "A", "0"))
tab_vowel[4,2] = coefficient_reports(filter_f0("B2", "A", "78"))
tab_vowel[4,3] = coefficient_reports(filter_f0("B2", "A", "90"))

tab_vowel[5,1] = coefficient_reports(filter_f0("C1", "A", "0"))
tab_vowel[5,2] = coefficient_reports(filter_f0("C1", "A", "78"))
tab_vowel[5,3] = coefficient_reports(filter_f0("C1", "A", "90"))

tab_vowel[6,1] = coefficient_reports(filter_f0("C2", "A", "0"))
tab_vowel[6,2] = coefficient_reports(filter_f0("C2", "A", "78"))
tab_vowel[6,3] = coefficient_reports(filter_f0("C2", "A", "90"))

tab_vowel[7,1] = coefficient_reports(filter_f0("D1", "A", "0"))
tab_vowel[7,2] = coefficient_reports(filter_f0("D1", "A", "78"))
tab_vowel[7,3] = coefficient_reports(filter_f0("D1", "A", "90"))

tab_vowel[8,1] = coefficient_reports(filter_f0("D2", "A", "0"))
tab_vowel[8,2] = coefficient_reports(filter_f0("D2", "A", "78"))
tab_vowel[8,3] = coefficient_reports(filter_f0("D2", "A", "90"))

tab_vowel[9,1] = coefficient_reports(filter_f0("A1", "E", "0"))
tab_vowel[9,2] = coefficient_reports(filter_f0("A1", "E", "78"))
tab_vowel[9,3] = coefficient_reports(filter_f0("A1", "E", "90"))

tab_vowel[10,1] = coefficient_reports(filter_f0("A2", "E", "0"))
tab_vowel[10,2] = coefficient_reports(filter_f0("A2", "E", "78"))
tab_vowel[10,3] = coefficient_reports(filter_f0("A2", "E", "90"))

tab_vowel[11,1] = coefficient_reports(filter_f0("B1", "E", "0"))
tab_vowel[11,2] = coefficient_reports(filter_f0("B1", "E", "78"))
tab_vowel[11,3] = coefficient_reports(filter_f0("B1", "E", "90"))

tab_vowel[12,1] = coefficient_reports(filter_f0("B2", "E", "0"))
tab_vowel[12,2] = coefficient_reports(filter_f0("B2", "E", "78"))
tab_vowel[12,3] = coefficient_reports(filter_f0("B2", "E", "90"))

tab_vowel[13,1] = coefficient_reports(filter_f0("C1", "E", "0"))
tab_vowel[13,2] = coefficient_reports(filter_f0("C1", "E", "78"))
tab_vowel[13,3] = coefficient_reports(filter_f0("C1", "E", "90"))

tab_vowel[14,1] = coefficient_reports(filter_f0("C2", "E", "0"))
tab_vowel[14,2] = coefficient_reports(filter_f0("C2", "E", "78"))
tab_vowel[14,3] = coefficient_reports(filter_f0("C2", "E", "90"))

tab_vowel[15,1] = coefficient_reports(filter_f0("D1", "E", "0"))
tab_vowel[15,2] = coefficient_reports(filter_f0("D1", "E", "78"))
tab_vowel[15,3] = coefficient_reports(filter_f0("D1", "E", "90"))

tab_vowel[16,1] = coefficient_reports(filter_f0("D2", "E", "0"))
tab_vowel[16,2] = coefficient_reports(filter_f0("D2", "E", "78"))
tab_vowel[16,3] = coefficient_reports(filter_f0("D2", "E", "90"))

tab_vowel[17,1] = coefficient_reports(filter_f0("A1", "U", "0"))
tab_vowel[17,2] = coefficient_reports(filter_f0("A1", "U", "78"))
tab_vowel[17,3] = coefficient_reports(filter_f0("A1", "U", "90"))

tab_vowel[18,1] = coefficient_reports(filter_f0("A2", "U", "0"))
tab_vowel[18,2] = coefficient_reports(filter_f0("A2", "U", "78"))
tab_vowel[18,3] = coefficient_reports(filter_f0("A2", "U", "90"))

tab_vowel[19,1] = coefficient_reports(filter_f0("B1", "U", "0"))
tab_vowel[19,2] = coefficient_reports(filter_f0("B1", "U", "78"))
tab_vowel[19,3] = coefficient_reports(filter_f0("B1", "U", "90"))

tab_vowel[20,1] = coefficient_reports(filter_f0("B2", "U", "0"))
tab_vowel[20,2] = coefficient_reports(filter_f0("B2", "U", "78"))
tab_vowel[20,3] = coefficient_reports(filter_f0("B2", "U", "90"))

tab_vowel[21,1] = coefficient_reports(filter_f0("C1", "U", "0"))
tab_vowel[21,2] = coefficient_reports(filter_f0("C1", "U", "78"))
tab_vowel[21,3] = coefficient_reports(filter_f0("C1", "U", "90"))

tab_vowel[22,1] = coefficient_reports(filter_f0("C2", "U", "0"))
tab_vowel[22,2] = coefficient_reports(filter_f0("C2", "U", "78"))
tab_vowel[22,3] = coefficient_reports(filter_f0("C2", "U", "90"))

tab_vowel[23,1] = coefficient_reports(filter_f0("D1", "U", "0"))
tab_vowel[23,2] = coefficient_reports(filter_f0("D1", "U", "78"))
tab_vowel[23,3] = coefficient_reports(filter_f0("D1", "U", "90"))

tab_vowel[24,1] = coefficient_reports(filter_f0("D2", "U", "0"))
tab_vowel[24,2] = coefficient_reports(filter_f0("D2", "U", "78"))
tab_vowel[24,3] = coefficient_reports(filter_f0("D2", "U", "90"))


tab_vowel

##            quiet    noise 78    noise 90
## A1 A -0.28786597 -0.24247753 -0.30183718
## A2 A -0.94189377 -1.21715432 -1.04074426
## B1 A  3.29930310  4.44473319  4.26599052
## B2 A -3.68320100 -5.22430946 -6.27214142
## C1 A -2.13631231 -2.06368135 -2.75352823
## C2 A  3.01033315  3.02702809  2.39415449
## D1 A  1.94669799  1.57254633  2.11046287
## D2 A -1.41100990 -3.51530530 -2.33342025
## A1 E -0.29252540 -0.84861170 -0.69958066
## A2 E -1.19115023 -1.48746256 -1.65442259
## B1 E  3.59328024  4.72965482  4.46913289
## B2 E -3.54701137 -5.91615578 -5.92709176
## C1 E -2.07185516 -2.16034262 -3.66593772
## C2 E  3.10283002  1.83602960  1.09639647
## D1 E  0.64208026  4.61889193  2.26070381
## D2 E -0.57781611 -2.13785431 -1.19070239
## A1 U -1.28368202 -0.94018817 -0.70906976
## A2 U -1.62964409 -1.85532732 -2.03872891
## B1 U  0.26794278  2.21307069  3.39750573
## B2 U -4.08087893 -6.07211141 -6.29807801
## C1 U -1.95566596 -3.13011073 -3.87637567
## C2 U  1.62192705  1.20632671 -0.03604034
## D1 U -2.58229256 -0.88871437 -1.66430780
## D2 U -3.99523644 -2.86434374 -2.10865905

F0 Analysis

Giang Le

11/05/2021